hIn the last few days, I’ve been struggling with the continuing inadequacy of AI, and the very real possibility that the teething troubles conversational technology is facing might boil down to something as simple as this;
That there’s a profound conflict between what makes a bot engaging, which is the human, creative writing, and what makes it capable of crunching and processing data, which is AI. The former, it seems to me, is being crushed by the constant developer-driven preference for technological solutions to non-technological problems. In this case, a blind faith in AI to take over the human role in a bot’s creation.
My recent reacquaintance with the much improved but still blissfully simple Landbot has only served to reinforce this sneaking, sacreligious view that AI might be more blocker than benefit to conversational tech right now. Landbot brings out the Hollywood Screenwriter in me. You can just let your creativity loose on this thing, and that free flow leads to far better work.
There are other huge benefits to Landbot’s “dumb bot” approach too. When you dispense with the AI, you remove the need for those laborious chatbot engines and their intents, entities and endless documentation, which means there’s no barrier to entry using this thing except the ability to write good dialogue and deliver a good structure. That’s great news for some of the very good writers I know, who struggle with the whole “designing” conversation bit, and would rather just write a conversation instead.
Landbot’s Keyword Jump feature gives the app depth and intelligence. A little more scale and versatility with an eye to enterprise clients would give it even more scope. Right now it’s a little “small biz” to be a powerhouse. But the foundations are rock solid so it’s not a stretch at all. Other premium features already provide most of what you need to build a better bot experience. It just needs to spread its wings.
Now, don’t get me wrong, you can’t dumb-bot your way out of every use case, even the majority of use cases. Most digital assistants will still require AI, but for formats and concepts that are outside the norm, that don’t require huge amounts of data crunching, can easily do without the fussy interjections of NLU, or the twisted algorithmic biases of NLG and open domain bots, a classic case of flawed solutionism.
Sure, any bot built on Landbot, or another non AI engine, will need to be carefully managed, adjusted and tweaked along the way, but that’s way easier with a “dumb bot” than it is with a supposedly “smart bot” dependent on AI and mounds of data. Dumb bots really are WYSIWYG and that can be very comforting.
The proof is in the pudding, of course. We need to take a complex, data-light chatbot brief and execute it using a stateful or stateless AI engine on the one hand, and Landbot on the other. My guess is that Landbot will deliver a more engaging, more dependable bot which engenders more user loyalty more of the time.
And if that turns out to be true, we can seriously let the creatives rip, and delight our users with good digital characters, humanised personality traits built on an empathic base, in a multiplicity of concepts and formats. All this without having to roll the contextual dice, because in a dumb bot the writer shapes the context.
And as we know, context is everything.